Imaging based cancer research is in the early stages of a radiomics revolution, as it has now become feasible to automatically extract large sets of quantitative image features relevant to prognosis or treatment response. In recent years, our research has attempted to tackle many of the computational and informatics challenges in associated with this field. This has included research and development in tools to help manage, integrate and analyze radiomic datasets. We are developing middleware systems for federating distributed sources of medical imaging data, radiation therapy data, and data derived from their human/automated analysis. These tools and systems have been adopted by nationally recognized data sources such as The Cancer Imaging Archive and in prominent studies such as the National Lung Screening Trial (NLST). We have also developed data management and analysis systems for digital pathology that have been funded and adopted by NCI for the NLST datasets and this work is currently being extended for the data collected as part of the Cancer Genome Atlas. We are now expanding these middleware systems with support for secure, on-demand cloud-driven storage provisioning, research data management and analysis.
The development of these computational tools and systems, while independent, has been strongly influenced by our informatics research activities. This has included in silico research experiments to explore the value of fusing complementary molecular, pathology and radiology data to better classify Glioblastomas. In collaboration with investigators from the Emory Transplant Center, we developed novel techniques and analysis algorithms for multi-dimensional flow cytometry to help characterize and quantify a transplant recipient’s immune repertoire, and explored its application in renal transplant patients. In a recent collaboration, we are targeting NSCLC patients with brain metastases. Here we are investigating the viability of fusing features extracted from advanced neuroimaging MR, with molecular data from NSCLC tumor site, to help stratify patients and predict therapies with the best outcome.